A minimal approach to RAG using Azure AI Search

🔍🤖💬 A minimal RAG (Retrieval-Augmented Generation) implementation that indexes FAQ data into Azure AI Search with vector embeddings and provides a chat terminal using Azure OpenAI for answering questions based on the retrieved content.

Setup

1. Copy .env.example → .env and add your Azure endpoints & keys
2. poetry install

Ingest FAQ

• To use Vector Search in Azure AI Search, a vectorizer can connect the Azure OpenAI Embedding model to turn your text into a numerical embedding. If you prefer, without using a vectorizer, you can also create the embedding yourself and provide it directly to Azure AI Search for Vector Search.
• Options 1 and 3 demonstrate how to provide the embedding directly without using a vectorizer. - chat_app.py uses VectorizedQuery.
• Options 2 and 4 demonstrate the use of a vectorizer along with semantic search. - chat_app_v2.py uses VectorizableTextQuery.

1. 📤Push - upload local data

   python push_aisearch_index.py 

2. 📤Push - upload local data w/ Vectorizer and Semantic Search

   python push_aisearch_index_v2.py 

3. 📥Pull - using Azure AI Search Indexer

   python pull_aisearch_index.py 

4. 📥Pull - using Azure AI Search Indexer w/ Vectorizer and Semantic Search

   python pull_aisearch_index_v2.py 

Chat

• To utilize an embedding directly with VectorizedQuery

  python chat_app.py 

• To utilize text with a vectorizer, use the VectorizableTextQuery method.

  python chat_app_v2.py 

• Type your question or exit to quit.

Azure AI Foundry

To connect Azure AI Search with Azure AI Foundry, you need to add both a vectorizer and semantic search to the index. Use an existing AI Search index with the Azure AI Search tool

📚 Learn more: csv indexer